Re: [ns] FTP Frustration

[Lloyd Wood <l.wood@eim.surrey.ac.uk>]

On Thu, 20 Dec 2001, Stanley M Bielski wrote:

> Hmmm... should it take over 40 seconds to send 4 packets of default size
> over 3.0 ms of link delay?

depends on link capacity (bandwidth), serialisation time (significant
if bandwidth is low) and competing traffic, in most cases. but not
yours.

> My networking knowledge is a little rusty, but if I tell FTP to send 40
> packets, shouldn't TCP send out the whole window? This is clearly not the
> case in my results. TCP sends out a packet, waits an geometrically
> increasing amount of time, and then sends out another.
>
> Any explanation for this behavior? Is LossMonitor not sending ACKs?

TCP sends out 1 packet/2 packets/4packets in increasing bursts
(flights) of packets. This is the slow-start process, and it's
well-documented in Stevens vol. 1, RFC2581, and elsewhere.

In ns, TCP sources work with TCP sinks. Lossmonitor == no TCP sink
== you never get beyond probing for an ack with the first packet,
because it never gets acked.
You're sending the first packet three times.

I can't see in the example code where you join the source to the sink
either, which would help. You create procedures but don't call them...

L.

looks just like my first ever ns tcl script, in fact.

<L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>